DEVELOPMENT OF MACHINE LEARNING MODEL TO DETECT DISEASES ON UBI CILEMBULEAVES (IPOMEA BATATAS (L). LAM)

• Main Author: Ari, Gerard
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/84061
• Institution: Institut Teknologi Bandung

In the last few years, Indonesia’s production on main sources of carbohydrates has been declining. This is also accompanied by the decrease in agricultural land, pest, and diseases that affects plant growth. In this study, a solution to combat these issues are introduced through smart agriculture that aims to increase efficiency and quality of plants which is represented through ubi Cilembu. This study is focused primarily on an early detection system to take care of pest and diseases that could be seen through the plant’s leaves. Two variation is made throughout the development of the machine learning model to obtain the most efficient model. The two variation made are on the training epoch and the image size. Throughout the development, a case of overfitting is found which is solved by doing data augmentation and increase in dataset. After doing this study, it is found that the most efficient training epoch is 100 epoch with an image size of 800x800 pixel. The model which is produced have a metric of 84.6% mAP, 81.3% precision, and 80.5% recall rate.